Hydroelectric generating set cluster simulation method

Abstract

The invention provides a hydroelectric generating set cluster simulation method, which comprises the following steps of: establishing a heterogeneous hydroelectric generating set independent model, calculating dynamic characteristics, and enabling a plurality of hydroelectric generating sets with different characteristics to be equivalent to a single set to be accessed into a power system simulation environment by adopting a cluster equivalence technology based on dynamic weight: S1, enabling a speed regulator of each set to be connected into a power system simulation environment according to power grid frequency deviation; the opening degree of the guide vane is calculated through PID control logic; s2, on the basis of a nonlinear water turbine model, the output torque of a single unit is calculated by combining the water head, flow, efficiency and rotating speed parameters of the unit; s3, determining a dynamic weight coefficient according to the unit capacity proportion and the real-time dynamic response factor, obtaining a cluster equivalent moment through weighted aggregation, inputting the cluster equivalent moment into an equivalent generator model, and calculating an equivalent rotating speed in combination with a rotor motion equation; and S4, single-machine dynamic characteristic independent calculation and cluster equivalent aggregation are realized through a decoupling aggregation architecture, and the analysis and calculation complexity of the power system is reduced while the simulation precision is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system simulation, in particular to a hydropower unit cluster simulation method. BACKGROUND

[0002] In traditional power system simulation, hydropower units are usually connected to the grid by using simplified models (such as constant power sources or first-order inertia models), but this method cannot accurately represent key dynamic characteristics such as water hammer effect of the water diversion system and nonlinearity of the governor. When multiple hydropower units are operating in parallel, they need to be modeled and coupled with the grid iteratively, resulting in an exponential increase in computational load. Although existing parallel simulation methods for hydropower units can improve computational efficiency, they do not solve the problems of model heterogeneity and cluster equivalence. In addition, the current dynamic aggregation technology is only applicable to the equivalence calculation of units of the same type, and does not fully consider the influence of differences in governor characteristics of different units on the dynamic response of the system, which to some extent restricts the accuracy of large-scale hydropower cluster simulation analysis.

[0003] The existing technology is a dynamic aggregation scheme that generates an equivalent single-machine model by performing arithmetic averaging on the inertia time constant and damping coefficient of multiple hydropower units. However, this averaging process ignores the differences in dynamic response caused by different lengths of the water diversion system, as well as the differences in control characteristics between mechanical hydraulic governors and digital electro-hydraulic governors. In addition, the equivalent model can only reflect the overall output characteristics of the cluster and cannot simulate the interaction phenomena such as frequency coupling and power oscillation between units. SUMMARY

[0004] The main purpose of the present application is to provide a hydropower unit cluster simulation method that achieves a balance between simulation accuracy and computational efficiency by establishing independent models of heterogeneous units, designing dynamic weight equivalence technology, and using a "decoupling aggregation" architecture.

[0005] To solve the above technical problems, the technical solution adopted by the present application is as follows: a hydropower unit cluster simulation method, which comprises the following steps: S1, each unit governor calculates the guide vane opening according to the grid frequency deviation through the PID control logic; S2, based on the nonlinear water turbine model, the output torque of a single unit is calculated by combining the unit head, flow rate, efficiency, and speed parameters; S3, the dynamic weight coefficient is determined according to the capacity ratio of the unit and the real-time dynamic response factor, the equivalent torque of the cluster is obtained by weighted aggregation, and the equivalent generator model is inputted, and the equivalent speed is calculated by combining the rotor motion equation; S4. By using a "decoupling and aggregation" architecture, independent calculation of dynamic characteristics of a single machine and equivalent aggregation of clusters are achieved, which reduces the complexity of power system analysis and calculation while ensuring simulation accuracy.

[0006] In the preferred scheme, the guide vane opening is calculated as follows: the governor of each unit is based on the grid frequency deviation. The guide vane opening is calculated using PID control logic. The calculation formula is: ; in , For reference frequency, The actual frequency of the power grid. , , These are the proportional coefficient, integral coefficient, and derivative coefficient of the speed governor, respectively. Unit torque calculation: Based on a nonlinear turbine model, the output torque of a single unit is calculated using unit parameters. The calculation formula is: ; in The density of water, It is the acceleration due to gravity. For the first Taiwanese turbine head, For the first Taiwan unit flow rate For the first Taiwanese unit efficiency. For the first Unit speed; Cluster equivalent moment calculation: Dynamic weighted aggregation: based on the dynamic weighting coefficients of the generating units. Through formula Computation cluster equivalent moment , This represents the total number of units within the cluster. Cluster access simulation: Equivalent moment Input the equivalent generator model and calculate the equivalent generator speed using the rotor motion equation. The rotor motion equation is , The equivalent inertial time constant, This is the equivalent damping coefficient; Dynamic weighting coefficient calculation: via formula Calculate dynamic weight coefficients ,in As an initial weight based on the proportion of unit capacity, This refers to the unit's real-time dynamic response factor. is an adjustment coefficient; The method adopts a "decoupling aggregation" architecture, independently calculates the dynamic characteristics of each heterogeneous unit through S1-S2, and realizes equivalent aggregation of the cluster through S3-S4, thereby reducing the complexity of power system analysis and calculation while ensuring simulation accuracy, and being suitable for power grid stability research and frequency regulation scenarios with large-scale hydropower units participating.

[0007] In the preferred scheme, when calculating the guide vane opening of the speed governor in step S1, the grid frequency data need to be preprocessed, and the specific process includes: A1. Frequency acquisition: Real-time acquisition of actual grid frequency through a high-frequency sampling module , the sampling frequency is set to 500Hz, and a continuous frequency data sequence is obtained ; A2. Filtering: Filtering the collected frequency data using Kalman filtering algorithm to eliminate high-frequency interference signals, and the state equation in the filtering process is set to , is the process noise, the observation equation is set to , is the observation value, and is the observation noise; A3. Deviation calculation: Substitute the filtered actual grid frequency and the reference frequency into to obtain an accurate frequency deviation value, and then input the PID control logic to calculate the guide vane opening .

[0008] In the preferred scheme, the determination of the parameters of the nonlinear water turbine model in step S2 needs to be realized through a working condition adaptation algorithm, and the specific steps include: B1. Working condition classification: According to the operating head range of the unit, the unit working condition is divided into low head working condition, medium head working condition and high head working condition; B2. Parameter retrieval: For different working conditions, retrieve the corresponding characteristic parameters from the unit parameter database, wherein the flow coefficient under low head working condition is , the efficiency coefficient is , the flow coefficient under medium head working condition is , the efficiency coefficient is , the flow coefficient under high head working condition is , and the efficiency coefficient is ; B3. Parameter calculation: Calculate the unit flow , by formula , wherein is the flow coefficient corresponding to the working condition, and calculate the unit efficiency , This represents the efficiency coefficient for the corresponding operating condition. The optimal head; B4. Torque Calculation: Calculate the torque... , With real-time data collection , Substitute into the formula The output torque of a single unit is obtained.

[0009] In the preferred scheme, the dynamic weight coefficient in dynamic weighted aggregation initial weights The calculation needs to be performed using a capacity occupancy analysis algorithm, and the specific steps include: C1. Capacity Acquisition: Acquire the rated capacity of each unit within the cluster. This forms a capacity data set. ; C2. Total Capacity Calculation: Calculated using the formula... Calculate the total capacity of the cluster ; C3. Initial Weight Determination: Based on the ratio of the capacity of a single unit to the total capacity, the initial weights are determined using the formula... Calculate the initial weights for each unit. This ensures that the initial weights reflect the basic contribution of unit capacity to cluster output.

[0010] In the preferred scheme, the equivalent parameters of the equivalent generator model in the cluster access simulation are... , The calculation needs to be performed using a parameter aggregation algorithm, and the specific steps include: D1. Single-unit parameter acquisition: Acquire the inertial time constant of each unit within the cluster. Damping coefficient ; D2. Weighted Calculation: Combining dynamic weight coefficients Through formula Calculate the equivalent inertial time constant Through formula Calculate the equivalent damping coefficient ; D3. Parameter Verification: Calculate the parameters... , Substituting the rotor motion equations, the speed response under small grid disturbances is simulated. If the deviation between the response curve and the average speed response curve of the units within the cluster is less than 5%, the parameters are considered valid; if the deviation is greater than or equal to 5%, the weighting coefficients are readjusted. Repeat steps D2D3 until the parameters meet the accuracy requirements.

[0011] In the preferred scheme, the dynamic response factor is used in the calculation of dynamic weighting coefficients. The calculation needs to be realized by a response characteristic analysis algorithm, and the specific steps include: E1. Response data collection: real-time collection of the speed change rate of each unit , guide vane opening change rate , and the sampling interval is set to 0.01s; E2. Normalization processing: the speed change rate is normalized by the formula , wherein is the maximum speed change rate of the unit, and the guide vane opening change rate is normalized by the formula , wherein is the maximum guide vane opening change rate of the unit; E3. Factor calculation: the dynamic response factor is calculated by the weighted formula , wherein the weight coefficients 0.6 and 0.4 are determined according to the influence degree of the speed response and the guide vane response on the dynamic characteristics of the unit.

[0012] In the preferred scheme, the determination of the adjustment coefficient in the dynamic weight coefficient calculation needs to be realized by a scene adaptation algorithm, and the specific steps include: F1. Scene classification: according to the grid operation scene, the simulation scene is divided into a steady-state scene, a disturbance scene and a fault scene; F2. Coefficient assignment: adjustment coefficients are set for different scenes, wherein under the steady-state scene, the weight adjustment range is reduced to ensure the simulation stability; under the disturbance scene, the weight adjustment sensitivity is moderately improved; under the fault scene, the weight adjustment effect is maximized to adapt to the severe dynamic change; F3. Scene switching: the grid frequency fluctuation is monitored in real time during the simulation process, and when the scene is switched, the adjustment coefficient is automatically updated to ensure that the dynamic weight coefficient can adapt to the current grid operation scene.

[0013] In the preferred scheme, the implementation of the “decoupling aggregation” architecture needs to be completed by a timing control algorithm, and the specific steps include: G1. Period division: the total simulation period is divided into a single-machine decoupling simulation period and a cluster aggregation period , and , wherein accounts for 70%, and accounts for 30%; G2. Decoupling simulation: in , each unit independently executes the steps S1-S2 to calculate the guide vane opening​ with output torque , each unit does not interact with data, avoiding coupling interference; G3. Data upload: at the end of , upload the , data of all units to the cluster control module; G4. Cluster aggregation: within , the cluster control module performs steps S3-S4 to complete the cluster equivalent calculation, generates equivalent parameters and accesses the power system simulation environment; G5. Cycle execution: repeat steps G2G4 to realize multi-cycle decoupling aggregation simulation, ensure continuous and stable simulation process.

[0014] In the preferred scheme, for large-scale hydropower unit clusters, the units need to be grouped first through a cluster grouping algorithm, including: H1. Characteristic index extraction: extract the key characteristic indexes of each unit, including governor type, rated head , rated capacity ; H2. Index standardization: standardize the rated head by formula , and standardize the rated capacity by formula ; H3. Similarity calculation: calculate the similarity of any two units using the Euclidean distance formula , 、 is the governor type identifier of the two units; H4. Grouping determination: set a similarity threshold , divide units with a similarity into the same group, execute the simulation steps of claim 1 in each group, and then aggregate the equivalent parameters of each group into the equivalent parameters of the entire cluster, reducing the calculation pressure of super large-scale clusters.

[0015] In the preferred scheme, the simulation process needs to be verified in real time by a precision verification algorithm to verify the simulation accuracy, including the following steps: I1. Actual data acquisition: real-time acquisition of actual total power , average speed of all units in the cluster through sensors; I2. Simulation data acquisition: read the equivalent power , equivalent speed output by the cluster equivalent model from the simulation system; I3. Error calculation: calculate the power error by formula , and calculate the speed error by formula ; I4. Precision Judgment: If and If the simulation accuracy meets the requirements, then the simulation accuracy is deemed to be satisfactory; if or Then return to step S4 to adjust the dynamic weight coefficients. Repeat steps S3-S4 until the error meets the accuracy requirements.

[0016] This invention provides a simulation method for hydropower unit clusters. By establishing a dynamic aggregation model of heterogeneous units based on multi-timescale decoupling, hydropower units with different characteristics are divided into several equivalent cluster units, and collaborative simulation with the main power grid is achieved through a dynamic interface matrix. This method effectively solves the contradiction between the loss of unit characteristics and low computational efficiency in traditional simulations, aiming to achieve synergistic optimization of simulation accuracy and efficiency when large-scale hydropower clusters participate in power grid stability analysis.

[0017] By using independent modeling of heterogeneous units (considering differences in governor type and head conditions) and dynamic weighting equivalence techniques, it is significantly superior to the traditional averaging aggregation method; By adopting a "decoupled aggregation" architecture and cluster equivalence technology, the computational workload is simplified from "exponential growth in the number of units" to "linear calculation of a single equivalent unit". The dynamic weighting coefficient can adapt to steady-state, disturbance, and fault scenarios of the power grid. The design of the adjustment coefficient and dynamic response factor makes the method applicable to different types and capacities of hydropower unit clusters. The generated equivalent parameters ( , It conforms to the input format of existing power system simulation tools and can be connected to mainstream simulation platforms such as PSASP and BPA without modification. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments: Fig. 1 This is a flowchart of the hydropower unit cluster simulation method of the present invention; Fig. 2 This is a diagram of the power system cluster access structure of the present invention; Fig. 3 This is a flowchart of the dynamic weight calculation sub-process of the present invention. Detailed Implementation

[0019] Example 1 like Figs. 1-3 As shown, a method for simulating a cluster of hydropower units includes: establishing independent models of heterogeneous hydropower units and calculating their dynamic characteristics; employing a cluster equivalence technique based on dynamic weights to equate multiple hydropower units with different characteristics to a single unit connected to a power system simulation environment; and including the following steps: S1. Each unit's speed governor calculates the guide vane opening based on the grid frequency deviation using PID control logic; S2. Based on the nonlinear turbine model, and combined with the unit's head, flow rate, efficiency, and speed parameters, calculate the output torque of a single unit; S3. Determine the dynamic weighting coefficients based on the unit capacity ratio and real-time dynamic response factors, obtain the cluster equivalent torque through weighted aggregation, input it into the equivalent generator model, and calculate the equivalent speed by combining the rotor motion equation. S4. By using a "decoupling and aggregation" architecture, independent calculation of dynamic characteristics of a single machine and equivalent aggregation of clusters are achieved, which reduces the complexity of power system analysis and calculation while ensuring simulation accuracy.

[0020] S1 guide vane opening calculation: The governor of each unit is based on the grid frequency deviation. The guide vane opening is calculated using PID control logic. The calculation formula is: ,in ( For reference frequency, (actual frequency of the power grid) , , These are the proportional coefficient, integral coefficient, and derivative coefficient of the speed governor, respectively. S2 Unit Torque Calculation: Based on a nonlinear turbine model and combined with unit parameters, the output torque of a single unit is calculated. The calculation formula is: ,in Let be the density of water, and g be the acceleration due to gravity. For the i-th generating unit, Let i be the flow rate of the i-th unit. Let be the efficiency of the i-th generating unit. Let be the rotational speed of the i-th unit; Calculation of equivalent moment of cluster in S3: S31. Dynamic weighted aggregation: Based on the dynamic weighting coefficient of the unit. Through formula Computation cluster equivalent moment (n is the total number of units in the cluster); S32. Cluster Access Simulation: Equivalent Moments Input the equivalent generator model and calculate the equivalent generator speed using the rotor motion equation. The rotor motion equation is ( The equivalent inertial time constant, (Equivalent damping coefficient); Calculation of dynamic weighting coefficients in S4: via formula Computing dynamic weight coefficients wherein is an initial weight based on the proportion of the unit capacity, is a real-time dynamic response factor of the unit, is an adjustment coefficient; The method adopts a "decoupling aggregation" architecture, first calculates the dynamic characteristics of each heterogeneous unit independently through S1-S2, and then realizes equivalent aggregation of the cluster through S3-S4, which reduces the complexity of power system analysis and calculation while ensuring simulation accuracy, and is suitable for large-scale hydropower unit participation in power grid stability research and frequency regulation scenarios.

[0021] The core logic of the hydropower unit cluster simulation method is to address the problem of difficult to balance heterogeneous hydropower unit characteristics representation and calculation efficiency in traditional power system simulation. Through a hierarchical design technical path, accurate equivalent and efficient simulation of multiple units are realized. First, in the single-machine dynamic characteristic calculation link, the governor of each unit will capture the deviation of the grid frequency and the rated reference frequency in real time, and dynamically calculate the guide vane opening with the help of the PID control logic (including proportional, integral, and differential adjustment functions). This process can accurately respond to changes in grid load, ensure the timeliness and stability of unit output regulation, and avoid the problem of non-linear characteristics of the governor that cannot be reflected in traditional simplified models. Then, based on the nonlinear water turbine model, the output torque of a single unit is calculated, fully considering the key parameters in the actual operation of the unit: water head (reflecting the potential energy of water flow), flow (directly related to the flow capacity of the guide vane opening), efficiency (dynamic energy conversion level with operating conditions), and speed (actual rotating state of the unit). Through multi-parameter coupling calculation, the unique output characteristics of different types of hydropower units (such as units equipped with mechanical hydraulic governors and digital electro-hydraulic governors) are fully restored, solving the defect of ignoring the heterogeneity of units in existing dynamic aggregation technology.

[0022] In the cluster equivalent link, the determination of the dynamic weight coefficient is the core innovation point. Instead of adopting fixed capacity proportion allocation, the dynamic weight coefficient is calculated by combining the rated capacity proportion of the unit (reflecting the static contribution difference) and the real-time dynamic response factor (reflecting the response speed of the unit to the power grid disturbance, the governing sensitivity of the guide vane, and other dynamic characteristics), so as to ensure that the weight of each unit can adapt to the change of the operating condition. On this basis, the torque of multiple units is integrated into the cluster equivalent torque through weighted aggregation, and then the equivalent generator model is input, and the equivalent speed is calculated by using the rotor motion equation. This step realizes the equivalent conversion from multiple units to a single unit, which not only retains the dynamic response characteristics of the cluster as a whole, but also avoids the problem of explosive calculation caused by traditional modeling one by one. Finally, the application of the decoupling and aggregation architecture divides the single-machine simulation and the cluster equivalence into two independent stages: in the single-machine stage, each unit calculates the dynamic characteristics independently without being directly coupled with other units or the power grid, thereby avoiding mutual interference in numerical calculation; in the aggregation stage, the single-machine data is integrated into equivalent parameters to be connected to the power grid simulation, thereby effectively solving the problem of numerical instability in the direct coupling simulation of large-scale units.

[0023] From the beneficial effects, the method first significantly improves the simulation accuracy. Through the independent modeling of heterogeneous units and the dynamic weight equivalence technology, the key information such as the governor characteristics and the difference of the water diversion system of different units can be completely retained, the loss of characteristics caused by the traditional simplified model or average aggregation is avoided, and the simulation result is more in line with the actual power grid operating state, thereby providing a more reliable decision basis for power grid stability research and frequency regulation analysis. Secondly, the calculation efficiency is greatly optimized. Through the cluster equivalence, the calculation amount of multiple units is simplified to the calculation of a single equivalent unit, and there is no need to iteratively couple each unit one by one, thereby effectively reducing the calculation complexity of power system analysis, especially suitable for large-scale cluster scenarios containing dozens of even hundreds of hydropower units, thereby providing feasibility for the simulation analysis of complex hydropower systems. In addition, the method has good adaptability and compatibility. The dynamic weight coefficient can be adaptively adjusted according to different operating scenarios such as power grid steady state, disturbance, and fault, and can be adapted to different types and capacities of hydropower units. At the same time, the generated equivalent parameters conform to the input format of the existing mainstream power system simulation tools, and do not need to modify the existing simulation platform, thereby directly connecting to the simulation platform, reducing the technical application threshold, and facilitating the popularization and use in actual engineering.

[0024] In the preferred scheme, the guide vane opening is calculated. The governor of each unit calculates the guide vane opening based on the frequency deviation of the power grid by the PID control logic , and the calculation formula is: ; wherein , is the reference frequency, is the actual frequency of the power grid, , , These are the proportional coefficient, integral coefficient, and derivative coefficient of the speed governor, respectively. Unit torque calculation: Based on a nonlinear turbine model, the output torque of a single unit is calculated using unit parameters. The calculation formula is: ; in The density of water, It is the acceleration due to gravity. For the first Taiwanese turbine head, For the first Taiwan unit flow rate For the first Taiwanese unit efficiency. For the first Unit speed; Cluster equivalent moment calculation: Dynamic weighted aggregation: based on the dynamic weighting coefficients of the generating units. Through formula Computation cluster equivalent moment , This represents the total number of units within the cluster. Cluster access simulation: Equivalent moment Input the equivalent generator model and calculate the equivalent generator speed using the rotor motion equation. The rotor motion equation is , The equivalent inertial time constant, This is the equivalent damping coefficient; Dynamic weighting coefficient calculation: via formula Calculate dynamic weight coefficients ,in As an initial weight based on the proportion of unit capacity, This refers to the unit's real-time dynamic response factor. For adjustment coefficients; This method adopts a "decoupled aggregation" architecture. First, the dynamic characteristics of each heterogeneous unit are calculated independently through S1-S2, and then the cluster is aggregated equally through S3-S4. While ensuring the simulation accuracy, it reduces the complexity of power system analysis and calculation. It is suitable for power grid stability research and frequency regulation scenarios involving large-scale hydropower units.

[0025] In the preferred embodiment, when the speed governor calculates the guide vane opening in step S1, it is necessary to preprocess the grid frequency data first. The specific process includes: A1. Frequency Acquisition: Real-time acquisition of the actual power grid frequency via a high-frequency sampling module. , the sampling frequency is set to 500Hz, and the continuous frequency data sequence is obtained ; A2. Filter processing: the collected frequency data is filtered by using Kalman filter algorithm to eliminate high-frequency interference signals, and in the filtering process, the state equation is set to , is the process noise, the observation equation is set to , is the observation value, is the observation noise; A3. Deviation calculation: the filtered actual frequency of the power grid and the reference frequency are substituted into to obtain the accurate frequency deviation value, and then the guide vane opening is calculated by inputting the PID control logic.

[0026] In the preferred scheme, the determination of the parameters of the non-linear hydraulic turbine model in step S2 needs to be realized by a working condition adaptation algorithm, and the specific steps include: B1. Working condition classification: according to the operating head range of the unit, the unit working condition is divided into low head working condition, medium head working condition and high head working condition; B2. Parameter retrieval: for different working conditions, the corresponding characteristic parameters are retrieved from the unit parameter database, wherein the flow coefficient under the low head working condition is , the efficiency coefficient is , the flow coefficient under the medium head working condition is , the efficiency coefficient is , the flow coefficient under the high head working condition is , and the efficiency coefficient is ; B3. Parameter calculation: the unit flow is calculated by the formula , is the flow coefficient corresponding to the working condition, and the unit efficiency is calculated by the formula , is the efficiency coefficient corresponding to the working condition, and is the optimal head; B4. Torque calculation: the calculated , and the real-time collected , are substituted into the formula to obtain the output torque of a single unit.

[0027] In the preferred scheme, the initial weight of the dynamic weight coefficient in the dynamic weighted aggregation needs to be calculated by a capacity proportion analysis algorithm, and the specific steps include: C1. Capacity Acquisition: Acquire the rated capacity of each unit within the cluster. This forms a capacity data set. ; C2. Total Capacity Calculation: Calculated using the formula... Calculate the total capacity of the cluster ; C3. Initial Weight Determination: Based on the ratio of the capacity of a single unit to the total capacity, the initial weights are determined using the formula... Calculate the initial weights for each unit. This ensures that the initial weights reflect the basic contribution of unit capacity to cluster output.

[0028] In the preferred scheme, the equivalent parameters of the equivalent generator model in the cluster access simulation are... , The calculation needs to be performed using a parameter aggregation algorithm, and the specific steps include: D1. Single-unit parameter acquisition: Acquire the inertial time constant of each unit within the cluster. Damping coefficient ; D2. Weighted Calculation: Combining dynamic weight coefficients Through formula Calculate the equivalent inertial time constant Through formula Calculate the equivalent damping coefficient ; D3. Parameter Verification: Calculate the parameters... , Substituting the rotor motion equations, the speed response under small grid disturbances is simulated. If the deviation between the response curve and the average speed response curve of the units within the cluster is less than 5%, the parameters are considered valid; if the deviation is greater than or equal to 5%, the weighting coefficients are readjusted. Repeat steps D2D3 until the parameters meet the accuracy requirements.

[0029] In the preferred scheme, the dynamic response factor is used in the calculation of dynamic weighting coefficients. The calculation needs to be implemented through a response characteristic analysis algorithm, and the specific steps include: E1. Response Data Acquisition: Real-time acquisition of the speed change rate of each unit. Guide vane opening change rate The sampling interval is set to 0.01s; E2. Normalization: This is achieved through the formula... Normalize the rate of change of rotational speed. The maximum rate of change of unit speed is expressed by the formula. Normalize the rate of change of guide vane opening. This represents the rate of change of the maximum guide vane opening of the unit. E3. Factor calculation: through the weighted formula Calculate the dynamic response factor , wherein the weight coefficients 0.6, 0.4 are determined according to the influence degree of the speed response and the guide vane response on the dynamic characteristics of the unit respectively.

[0030] Example 2 Adjustment coefficient in dynamic weight coefficient calculation The determination of the adjustment coefficient needs to be realized through a scene adaptation algorithm, and the specific steps include: F1. Scene classification: according to the grid operation scene, the simulation scene is divided into steady-state scene, disturbance scene and fault scene; F2. Coefficient assignment: set the adjustment coefficient for different scenes , wherein in the steady-state scene , the weight adjustment range is reduced to ensure the simulation stability; in the disturbance scene , the weight adjustment sensitivity is moderately improved; in the fault scene , the weight adjustment effect is maximized to adapt to the severe dynamic change; F3. Scene switching: real-time monitor the grid frequency fluctuation during the simulation process, when the scene is switched, automatically update the adjustment coefficient , ensure that the dynamic weight coefficient can adapt to the current grid operation scene.

[0031] In the preferred scheme, the implementation of the "decoupling aggregation" architecture needs to be completed through a timing control algorithm, and the specific steps include: G1. Period division: divide the total simulation period into single-machine decoupling simulation period and cluster aggregation period , and , wherein accounts for 70%, and accounts for 30%; G2. Decoupling simulation: in , each unit independently executes steps S1-S2 to calculate the guide vane opening and the output torque , and there is no data interaction between the units to avoid coupling interference; G3. Data uploading: at the end of , upload the , data of all units to the cluster control module; G4. Cluster aggregation: in , the cluster control module executes steps S3-S4 to complete the cluster equivalent calculation, generates equivalent parameters and accesses the power system simulation environment; G5. Loop execution: repeat G2G4 steps to realize multi-cycle decoupling aggregation simulation, ensure continuous and stable simulation process.

[0032] In the preferred embodiment, when applied to large-scale hydropower unit clusters, the units need to be grouped through a cluster grouping algorithm, including: H1. Characteristic index extraction: extract the key characteristic indexes of each unit, including governor type, rated head , rated capacity ; H2. Index standardization: standardize the rated head by formula , and the rated capacity by formula ; H3. Similarity calculation: calculate the similarity of any two units using the Euclidean distance formula , 、 is the governor type identifier of the two units. H4. Grouping determination: set a similarity threshold , and divide units with a similarity into the same group. The simulation steps in claim 1 are performed within each group, and the equivalent parameters of each group are aggregated into the equivalent parameters of the entire cluster, reducing the computational pressure of super-large clusters.

[0033] In the preferred embodiment, the simulation process needs to be verified in real time by a precision verification algorithm. The specific steps include: I1. Actual data acquisition: real-time acquisition of actual total power , average speed of all units in the cluster through sensors; I2. Simulation data acquisition: read the equivalent power , equivalent speed output by the cluster equivalent model from the simulation system; I3. Error calculation: calculate the power error by formula , and the speed error by formula ; I4. Precision judgment: if and , it is determined that the simulation precision meets the requirements; if or , return to step S4 to adjust the dynamic weight coefficient , and re-execute steps S3-S4 until the error meets the precision requirements.

[0034] I. Scene adaptation algorithm and formula correlation of adjustment coefficient a The adjustment coefficient a is a key parameter for calculating the dynamic weight coefficient ki. Its core function is to dynamically adjust the influence of the dynamic response factor kd on the weight according to the differences in the grid operation scenarios, to ensure that ki can adapt to the current grid state and ensure the accuracy of the cluster equivalence. This parameter is determined by the scenario adaptation algorithm, and the association of each step of the algorithm with the data is as follows: (1) Scenario classification (F1): Scenario definition based on grid frequency fluctuation The core basis for scenario classification is the deviation of the actual grid frequency from the reference frequency , which directly reflects the stability of the grid operation: When , the grid frequency fluctuation is minimal, and the load and output are basically balanced. At this time, there is no need to frequently adjust the weight, and the simulation stability needs to be prioritized. When , the grid experiences small load fluctuations (such as short-term changes in industrial load), and the frequency is in a state of slow change. The weight adjustment sensitivity needs to be moderately improved to follow the dynamic changes. When , the grid may experience short-circuit faults, large-capacity load switching, and other faults, and the frequency fluctuates dramatically. The weight adjustment effect needs to be maximized to quickly adapt to extreme dynamic characteristics.

[0035] The data comes from real-time collection and filtering of grid frequency (same as the frequency deviation data calculated in step S1), and through the analysis of the fluctuation range of the same set of frequency data, the automatic division of scenarios is realized, ensuring the data consistency of scenario classification and unit control logic.

[0036] (2) Coefficient assignment (F2): Correspondence between scenarios and a and data significance The values set for different scenarios (stable scenario , disturbance scenario , fault scenario ) are not fixed empirical values, but are derived based on the matching of the calculation formula of the dynamic weight coefficient ( ) and the grid scenario requirements: Stable scenario : At this time, the adjustment range of to is , and the weight is mainly determined by the initial capacity weight ​The decision was made to retain only a small dynamic adjustment margin. This is because the dynamic response differences of the unit are small under steady-state conditions, and excessive adjustment would lead to weight fluctuations, affecting the stability of the simulation results; here... The optimal value was obtained through multiple steady-state simulations—when At times, the weights cannot respond to minute dynamic changes; when At that time, fluctuations in weights can cause the equivalent moment calculation error to exceed 1%, which does not meet the accuracy requirements.

[0037] Disturbance scene The adjustment range has been increased to ,make It can retain the fundamental role of capacity weights, and also... This reflects the difference in the response speed of the generating units. For example, when two generating units of the same capacity ( In the same case, the rate of change of guide vane opening of one guide vane is twice that of the other. Larger) This can increase the weight of units with fast response times. This accurately reflects the differences in their contributions to the cluster's output, avoiding equivalence errors caused by weight equalization.

[0038] Fault Scenario : Adjustment range maximized to ,at this time This becomes the dominant influencing factor in weighting. Under fault conditions, the unit's dynamic response capability (such as the rate of change of speed and guide vane adjustment speed) directly determines the frequency recovery speed of the cluster, through significant... The value allows for higher weighting of fast-responding units, increasing their share of the cluster's equivalent torque. The equivalent model can more realistically simulate the cluster's rapid response to faults, avoiding the shortcomings of traditional average weighting which fails to reflect the "role of key units".

[0039] (III) Scene Switching (F3): Real-time Updates and Data Linkage of α The core of scene switching is based on Real-time monitoring data to achieve The automatic update and its data linkage logic are as follows: Monitoring frequency: Use the same sampling frequency (500Hz) as in step S1, and calculate in real time. The instantaneous value and sliding average value (sliding window set to 0.02s) are used to avoid misjudgment of the scene due to fluctuations in instantaneous frequency; Switch trigger: after moving average If the current scene threshold is exceeded for three consecutive sampling periods (0.006s), a scene switch is triggered—for example, from a steady-state scene ( Switch to the disturbed scene ( ) is met the moving average of 0.006s is in the interval of 0.02Hz~0.1Hz; Data synchronization: After updating, it is immediately synchronized to the dynamic weight coefficient calculation module to recalculate , and the new is substituted into the cluster equivalent moment calculation of step S3 to ensure that the equivalent parameters are synchronized and adapted to the current scene. For example, when the power grid switches from a disturbance scene to a fault scene, from 0.5 to 0.8, the influence ratio of increases, the cluster equivalent moment will tilt faster to the torque of the unit with fast response, and the equivalent speed calculation can reflect the frequency recovery trend of the cluster in real time, avoiding simulation deviation caused by scene lag.

[0040] II. Timing control algorithm and formula association of "decoupling-aggregation" architecture The "decoupling-aggregation" architecture divides the simulation period through the timing control algorithm to realize the orderly connection of single machine independent simulation and cluster equivalence. The core is the determination of the period division ratio (70% for and 30% for ) and the flow of data in each stage. The specific formula and data association are as follows: (I) Period division (G1): setting of the ratio of T1 and T2 and data basis The setting of the total simulation period needs to match the time step requirement of power system simulation (usually 0.01s~0.05s). In this application, the default (considering simulation accuracy and calculation efficiency), where (70%) and (30%), the determination of this ratio is based on the following data logic: The duration of: single machine simulation (S1-S2 steps) needs to complete three core processes of frequency filtering, PID calculation, and torque calculation. The calculation time of each process is obtained through actual measurement: frequency Kalman filter takes about 0.003s, PID integral calculation takes about 0.004s, and nonlinear hydraulic turbine torque calculation takes about 0.005s. The total time is about 0.012s, with a redundant time of 0.002s (to cope with data transmission delay), so 0.014s (70%) is set to ensure that single machine simulation can be completed within the independent period, avoiding time conflicts with other units or cluster calculations; Duration basis: cluster aggregation (S3-S4 steps) needs to complete weight calculation, moment aggregation, equivalent speed calculation, and the calculation time is about 0.004s: dynamic weight coefficient calculation (including adaptation) takes 0.001s, equivalent moment weighted summation takes 0.001s, and rotor motion equation integral calculation takes 0.002s, and 0.002s of redundant time is reserved (to cope with multi-unit data synchronization), so 0.006s (30%) is set to ensure that the cluster equivalent can be completed in a short time without affecting the real-time performance of the power system simulation.

[0041] Period division formula It is not simply time superposition, but through time slice allocation to realize the logical isolation of "decoupling" and "aggregation": Each unit calculates independently, and data is only cached in the machine without interaction with the outside; Only the cluster control module works, and the cached data of all units is called for aggregation, avoiding the numerical instability problem caused by the simultaneous "calculation-interaction" in traditional parallel simulation.

[0042] (2) Decoupled simulation (G2): single machine data independent calculation and cache logic In , each unit independently executes S1-S2 steps, and its data independence is reflected in the following aspects: Input data independence: Although the frequency data of each unit comes from the same power grid, it needs to be processed separately by the filter module of the machine (to avoid delay caused by data sharing), for example, unit 1 and unit 2 collect at the same time, but the filter result of unit 1 is 49.948Hz, and that of unit 2 is 49.952Hz, which are used for the guide vane opening calculation of each unit respectively, reflecting the slight difference of different unit sensors; Independent calculation process: PID parameters (Kp, Ki, Kd) , , ) are retrieved separately according to the type of unit governor (mechanical and hydraulic governor parameters are different from digital electro-hydraulic governor parameters), and the water head and speed in moment calculation are also real-time data collected for each unit, for example, the water head of unit A and the water head of unit B are substituted into the nonlinear hydraulic turbine model to calculate the moment, ensuring that the characteristics of heterogeneous units are not averaged; Data cache independence: the , After the calculation is completed, the data is stored in the local cache unit. The cache unit is set with a timestamp (accurate to 1μs) to ensure that subsequent data uploads are aligned in time order and to avoid data timing disorder caused by differences in calculation speed.

[0043] (III) Data Upload (G3): Standalone Data Synchronization and Verification Logic At the end, the data flow switches to a "centralized upload" mode, and its data association logic is as follows: Upload trigger: When the cluster control module receives the "computation complete" signal from all units (or When the countdown ends, a data upload command is triggered, carrying a unified timestamp (e.g., ...). ), ensuring that all units begin uploading at the same time; Data verification: The data uploaded by each unit includes , and the corresponding calculation input data ( , , The cluster control module performs verification through "reverse verification"—for example, based on the uploaded... Recalculate with PID parameters If it is the same as the uploaded If the deviation exceeds 0.1%, the data is considered abnormal, and the unit is required to re-upload the data to prevent erroneous data from entering the aggregation process. Data alignment: Uploaded data from all units is sorted by timestamp to ensure that single-unit data within the same simulation cycle is grouped together. For example, all unit data with timestamp $0.014s$ is marked as "Single-unit data for cycle 1" for subsequent simulations. Aggregate computing provides support for data groups.

[0044] (iv) Cluster Aggregation (G4): Equivalent Parameter Calculation and Grid Connection Logic exist Within the cluster control module, steps S3-S4 are executed based on the uploaded single-machine data. The relationship between the formula and the data is as follows: Dynamic weight calculation: This is the output of the scenario adaptation algorithm. Value, combined with the uploaded unit capacity (calculate ), speed change rate (calculate ),pass Calculate each unit , here These are inherent parameters of the generator set (retrieved from the database). Uploaded Data difference calculation yields ( , is the rotational speed data of the last period); Equivalent moment aggregation: aggregate the uploaded with the calculated Substitute , is the total number of units in the cluster (automatically confirmed according to the number of units uploading data, to avoid calculation errors caused by unit shutdown), for example, if 2 of the 32 units are shut down (no data uploaded), then , only the of 30 units is summed up; Equivalent speed calculation: according to the aggregated , combined with the equivalent parameters , (calculated by , , , are inherent parameters of the unit), substitute into the rotor motion equation , calculate by Euler integration (the integration step is consistent with , which is 0.02s); Grid access: encapsulate the calculated , into a data frame that meets the input format of power system simulation tools (such as PSASP), and transmit it into the grid simulation environment through the standard interface. The data frame carries a timestamp (such as ), which ensures synchronization with the time step of the grid simulation.

[0045] (Five) Cycle execution (G5): Multi-cycle data flow and simulation continuity The cycle execution of G2-G4 steps is to realize the continuity of simulation through "cycle counter" and "data iteration". The data association logic is as follows: Cycle counting: the cluster control module sets the cycle counter, which starts from 1 and increments by 1 every time a ( ) counter is added by 1, and the counter value is used as the period identifier of the data frame, for example, the timestamp of the 5th period is ; Data iteration: the data of the last period is used as the initial value of the equivalent speed calculation of the next period ( , is the current period number), which ensures the continuous change of equivalent speed without jump; at the same time, the data of the last period is used as the initial value of the weight calculation of the next period, if the scene is not switched, only according to Small changes in the micro update, avoid weight fluctuations caused by simulation shock; Abnormal processing: if part of the unit data upload fails in a certain period, the cluster control module will use the last period's data of the unit , and mark it as "estimated data", while reducing the unit's (multiply by a decay factor of 0.5) to reduce the impact of estimated data on the aggregation result. When the unit data returns to normal, the decay factor is automatically removed to ensure that the simulation process does not stop.

[0046] Three, grouping algorithm and formula association of large-scale cluster When the number of units exceeds 50 (large-scale cluster), direct equivalent calculation of all units will result in excessive data processing, so the cluster grouping algorithm is needed to divide the units into several groups, each group is independently equivalent and then aggregated as a whole. The formula and data association are as follows: (1) Characteristic index extraction (H1): parameter selection and data significance for grouping The key characteristic indexes (governor type , rated head , rated capacity ) are core parameters that determine the dynamic characteristics of the unit. Their data sources and significance are as follows: Governor type : a discrete identification data, mechanical hydraulic governor is marked as 1, digital electro-hydraulic governor is marked as 2. This parameter directly affects the guide vane regulation speed of the unit (digital electro-hydraulic governor usually has 2-3 times the speed of mechanical hydraulic governor), and is the primary basis for grouping - if units with two types of governors are mixed into one group, the dynamic response difference between units in the group will be too large, and the equivalent error will exceed 5%; Rated head : a unit-specific parameter (retrieved from the database, unit: m), reflecting the design and operation head range of the unit. Units with similar rated heads have more similar flow and efficiency characteristics at the same guide vane opening (for example, units with rated heads of 100 m and 105 m have an efficiency deviation of less than 2% at an actual head of 102 m). If units with a rated head difference of more than 20 m are grouped together, the non-linear hydraulic turbine model parameter difference will be too large, and the moment calculation error will increase significantly; Rated capacity : the inherent parameters of the unit (retrieved from the database, unit: MW), reflecting the output scale of the unit. The output adjustment range of units with similar capacity under grid disturbance is more matched (for example, the maximum output deviation of 700 MW unit and 600 MW unit is about 14%, while the maximum output deviation of 700 MW unit and 300 MW unit is up to 133%). The capacity difference will lead to an imbalance in the contribution of the unit to the cluster output, and the equivalent model cannot reflect the actual output distribution within the group.

[0047] (II) Index standardization (H2): formula and data processing to eliminate dimensional influence Since the rated water head (unit: m) and the rated capacity (unit: MW) have different dimensions, directly calculating the similarity will lead to the dominance of parameters with large dimensions, which need to be converted to dimensionless data between 0 and 1 through the standardization formula. The formula and data correlation are as follows: The rated water head standardization formula : , are the minimum and maximum values of the rated water head of all units in the cluster (calculated by traversing the data of all units), for example, the rated water head of units in the cluster ranges from 80 m to 120 m, then , ; If the of a unit is , then , the standardized data reflects the relative position of the unit's water head in the cluster, avoiding the weight imbalance caused by absolute value difference (such as 80 m and 120 m).

[0048] The rated capacity standardization formula : are the minimum and maximum values of the rated capacity of all units in the cluster (calculated by traversing the data of all units), for example, the rated capacity of units in the cluster ranges from 300 MW to 700 MW, then , ; If the of a unit is , then , the standardized data is consistent in dimension and can participate in similarity calculation together.

[0049] In the standardization process, if the of all units are the same (such as the same type of units in the same hydropower station), then , in which case (Avoiding the denominator is zero), the same applies to The same case, to ensure that the formula can still be calculated in special cases.

[0050] (Three) similarity calculation (H3): Euclidean distance formula and characteristic difference quantification Using the Euclidean distance formula to calculate the similarity of the characteristics between the units Its core is to convert the difference of three-dimensional characteristic indicators ( , , ) into one-dimensional distance value, and the formula is associated with the data as follows: The physical meaning of each term in the formula: : Reflects the difference in governor type. If the governor types of the two units are the same ( ), this term is 0; if the types are different (one is 1 and the other is 2), this term is 1, which is the largest weight among all indicators, ensuring that units with different governor types will not be divided into the same group. : Reflects the difference in normalized head. For example, if the of unit A is 100 and the of unit B is 99, this term is 0.01. The smaller the difference, the smaller the value of this term. : Reflects the difference in normalized capacity. For example, if the of unit A is 100 and the of unit B is 99, this term is 0.01. Similarly, it reflects the square relationship of the difference to avoid offsetting positive and negative differences.

[0051] The meaning of the distance value: The value range of is 0~√3 (about 1.732), The smaller the value, the more similar the characteristics of the two units. For example, if the governor types of the two units are the same ( ), , , the characteristics are completely identical. If the governor types of the two units are different, the difference is 0.5, and the difference in is 0.5, the characteristics are significantly different.

[0052] (Four) grouping determination (H4): threshold setting and in-group equivalence logic The setting of the similarity threshold and the in-group equivalence logic directly determine the grouping effect and the calculation complexity, and their data association is as follows: Determination of the threshold : Through multiple simulation verifications, when At that time, the dynamic response difference between units within the group can be controlled within 10%, and the equivalent error is less than 3% (meeting accuracy requirements). If If there are too many groups (e.g., 50 machines divided into 20 groups), although it can reduce intra-group error, it increases the computational load of inter-group aggregation; if If the number of groups is too small (e.g., 50 machines are divided into 5 groups), the error within each group will exceed 5%, which does not meet the accuracy requirements. Grouping process: First, select one ungrouped unit as the "seed unit" and calculate its relationship with all other ungrouped units. ,Will The units are grouped together; then new seed units are selected from the remaining ungrouped units, and the above process is repeated until all units are grouped. Inter-group aggregation: Within each group, the simulation steps (S1-S4) described in claim 1 are executed to generate the equivalent moments for each group. , ... ( (Number of groups) and equivalent speed , ... Then, based on the proportion of the total capacity of each group ( ), calculate the equivalent moment of the entire cluster. With equivalent speed This enables hierarchical equivalence for ultra-large-scale clusters, reducing computational cost compared to direct equivalence. (like (At that time, the computational workload was reduced by 90%).

[0053] IV. Simulation Accuracy Verification Algorithm and Formula Correlation The accuracy verification algorithm achieves real-time monitoring and dynamic correction of simulation accuracy by comparing the error between simulation data and actual data. Its core lies in the design of the error calculation formula and the accuracy judgment logic, and its specific relationship with the data is as follows: (I) Actual Data Acquisition (I1): Acquisition and Processing of Real-World Operational Data Actual total power collected With average speed This is the baseline data for error calculation, and its acquisition and processing logic is as follows: Actual total power The actual output power of each unit is collected in real time by a power sensor at the unit outlet. (The sampling frequency is consistent with the simulation period, which is 0.02s), through To calculate the total power of the cluster, the collected data needs to be low-pass filtered (cutoff frequency 5Hz) to eliminate high-frequency noise in power fluctuations (such as instantaneous power jumps caused by electromagnetic interference). Average speed : Actual speed of each unit is collected by speed sensor of main shaft of the unit (Sampling frequency: 500 Hz), through The average speed of the cluster is calculated by arithmetic mean rather than weighted mean, because the speed is a direct reflection of the grid frequency (all unit speeds are synchronized with the grid frequency), and the average speed can more truly represent the speed state of the cluster.

[0054] The actual data collection needs to be strictly aligned with the timestamp of the simulation data, for example, the timestamp of the simulation cycle is , then the actual data needs to collect the average value within 0.005s before and after this time point, to ensure consistency in time dimension, and avoid errors caused by time deviation.

[0055] (II) Simulation data acquisition (I2): extraction logic of equivalent model output data The equivalent power and equivalent speed read from the simulation system are the core outputs of the cluster equivalent model, and their data extraction and correlation are as follows: Equivalent power : According to the equivalent speed and equivalent torque , the equivalent power is calculated by the basic formula of power calculation of power system (1.05 is the power correction coefficient, used to compensate for the mechanical loss and electrical loss ignored in the equivalent model, this coefficient is determined by actual measurement - in actual operation, the output power of the unit is about 95%~98% of the mechanical power, so take 1.05 as the correction coefficient, so that is closer to the actual power); Equivalent speed : Directly extracted from the output of the cluster equivalent module, i.e. calculated by the rotor motion equation, this data is bound to the timestamp of the simulation cycle, and can be directly compared with the actual average speed .

[0056] The extraction of simulation data needs to ensure consistency with the calculation logic of actual data, for example, The calculation is based on the correction of mechanical power , while is the actual output electric power, the physical meaning of the two is matched, avoiding "invalid error" caused by difference in calculation logic.

[0057] (III) Error calculation (I3): relative error formula and precision quantification Relative error is used instead of absolute error because the absolute values ​​of power and rotational speed differ greatly between clusters of different sizes (e.g., the absolute error ranges are different for a 100MW cluster and a 10000MW cluster). Relative error can more objectively reflect the simulation accuracy. The relationship between the formula and the data is as follows: Power relative error formula : molecular The denominator represents the absolute deviation between the simulated power and the actual power. The relative error eliminates the influence of the absolute power value, representing the actual power. For example, the relative error of 5MW for a 100MW cluster and 500MW for a 10000MW cluster is 5%, both reflecting the same simulation accuracy. like (If all units are shut down), then the default is... (At this point, the simulated power should also be 0, with no error), to avoid the denominator being zero.

[0058] Formula for relative error of rotational speed : The reference value for rotational speed is the actual average rotational speed. (Typically close to the rated speed of 104.72 rad / s), the relative error reflects the proportion of deviation in the speed simulation. For example, , The absolute deviation is 0.21 rad / s, and the relative error is... It reflects extremely high accuracy in speed simulation.

[0059] (iv) Accuracy Judgment (I4): Error Threshold Setting and Correction Logic Error threshold ( , The setting is based on the engineering requirements of power system simulation, and its judgment and correction logic is as follows: Engineering basis for the threshold: In power grid stability studies, a power simulation error exceeding 3% can lead to misjudgment of the unit's frequency regulation capability (e.g., if the actual frequency regulation capacity is 100MW, a simulation value below 97MW will be considered as insufficient frequency regulation capability), and a speed error exceeding 2% can lead to deviations in frequency stability analysis (e.g., if the actual frequency is 50Hz, a simulation value below 49Hz will be misjudged as a risk of frequency collapse). Therefore, this threshold is set to ensure that the simulation results can support engineering decisions. Correct the trigger logic: If or If the simulation accuracy does not meet the requirements, it is necessary to return to step S4 to adjust the dynamic weighting coefficients. The adjustment logic is as follows: For units that cause errors (such as those with the largest deviation between simulated and actual power values), correct them according to the direction of the error. (If the simulated power is too low, increase the power output of the unit with the fastest response time.) (increase its weight), recalculate Then, execute steps S3-S4 until the error meets the threshold requirement; Correction effect verification: After each correction, the error needs to be monitored for 3 consecutive simulation cycles. If the threshold requirements are met, the correction should be stopped. If the requirements are not met, the correction process should be repeated, up to 5 times (to avoid infinite loops). If the requirements are still not met after 5 times, the data acquisition should be checked for abnormalities (such as sensor failure) to ensure that the simulation accuracy problem can be located and resolved in a timely manner.

[0060] Example 3 Further explanation in conjunction with Example 1, such as Figs. 1-3 As shown in the diagram, this invention achieves cluster simulation of hydropower units through a four-step core process: guide vane opening calculation (S1), unit torque calculation (S2), cluster equivalent torque calculation (S3), and dynamic weight coefficient calculation (S4). It adopts a "decoupling aggregation" architecture: first, the dynamic characteristics of each heterogeneous unit are calculated independently, and then multiple units are equivalent to a single unit and connected to the power system simulation environment through dynamic weight equivalence technology. This not only ensures the simulation accuracy but also significantly reduces the computational complexity.

[0061] (I) Guide vane opening calculation (S1): Opening adjustment based on PID control logic The guide vane opening is a core control parameter for hydropower units in response to changes in grid frequency. It needs to be dynamically calculated based on the grid frequency deviation to ensure that the unit output matches the grid load demand.

[0062] 1. Frequency Deviation Calculation formula: ; Symbol definition: Power grid frequency deviation (unit: Hz). The reference frequency for the power grid is typically taken as the rated frequency of 50Hz or 60Hz. In this application, 50Hz is used by default as the fixed input frequency. The actual frequency of the power grid, in Hz, is collected in real time via a high-frequency sampling module at a sampling frequency of 500Hz. The collected data is a continuous time series. High-frequency interference needs to be eliminated by Kalman filtering.

[0063] Data association: This is the standard rated frequency of the power system, serving as the benchmark data for simulation. The difference between the real-time collected power grid operation data and the power grid load change status is... It is the core basis for triggering the unit's speed governor to operate—when When the grid frequency is higher than the rated value, the guide vane opening needs to be reduced to decrease the unit output; when When the grid frequency is lower than the rated value, the guide vane opening needs to be increased to increase the unit output.

[0064] 2. PID control logic calculates the guide vane opening Formula: ; Symbol definition: is the guide vane opening change, unit: %, value range is 10%~+10%, corresponding to the closing and opening of the guide vane, is the proportional coefficient, unitless, determined according to the type of unit governor: mechanical hydraulic governor take 0.8~1.2, digital electro-hydraulic governor take 1.0~1.5, is the integral coefficient, unit: 1 / s, mechanical hydraulic governor take 0.1~0.3, digital electro-hydraulic governor take 0.2~0.4, is the differential coefficient, unit: s, mechanical hydraulic governor take 0.05~0.15, digital electro-hydraulic governor take 0.1~0.2; is the integral term, the integral of the frequency deviation with respect to time, which eliminates the static error, is the rate of change of frequency deviation, unit: Hz / s, reflects the trend of frequency change, adjusts the guide vane opening in advance to suppress the deviation from expanding.

[0065] Data association: , , is the inherent parameter of the governor, which needs to be retrieved from the unit parameter database according to the type of the actual configured governor, and is the key preset data to ensure the accuracy of PID control; is the real-time calculated frequency deviation data, which directly determines the input value of the PID three terms: proportional term quickly responds to the deviation, integral term eliminates long-term deviation, differential term predicts the trend of deviation; The calculated is the adjustment instruction of the guide vane opening, which is directly used as the input data for subsequent unit torque calculation, and is the core control logic related to unit output regulation.

[0066] (2) Unit torque calculation (S2): output representation based on nonlinear water turbine model The unit output torque is the core parameter reflecting the mechanical power of the unit, which needs to be calculated in combination with the nonlinear characteristics of the water turbine (the relationship between water head, flow, efficiency and speed) to ensure accurate representation of the output capacity of the unit under different working conditions.

[0067] Formula: ; Symbol definition: is the output torque of the first unit, unit: N·m, negative sign indicates that the torque direction is opposite to the unit rotation direction, which hinders the unit acceleration, consistent with the actual physical characteristics; is the density of water, a fixed physical parameter, taking 1000 kg / m³, is the acceleration of gravity, a fixed physical parameter, taking 9.81 m / s²; is the water head of the first unit, unit: m, real-time data acquisition, calculated according to the difference between the reservoir water level and the downstream tail water level, divided into three working conditions: low water head, <80 m, medium water head, 80~120 m, high water head, >120 m; is the flow of the first unit, unit: m³ / s, calculated according to the guide vane opening and the water head , the formula is , is the flow coefficient, low water head working condition takes 0.8~1.0, medium water head working condition takes 1.0~1.2, high water head working condition takes 1.2~1.4; is the efficiency of the first unit, unitless, value range 0~1, calculated according to the water head and the optimal water head .

[0068] The formula is , is the efficiency coefficient, low water head working condition takes 0.85~0.90, medium water head working condition takes 0.90~0.95, high water head working condition takes 0.88~0.93, is the optimal water head of the unit, retrieved from the unit parameter database; is the rotation speed of the first unit, unit: rad / s, real-time data acquisition, rated rotation speed corresponds to 50 Hz, which is rad / s, about 104.72 rad / s.

[0069] Data association: , is a fixed physical constant, which is the benchmark data for torque calculation and does not need to be adjusted in real time; For real-time acquisition of water head data, directly affect the flow With efficiency : the higher the water head, the greater the potential energy of water flow, the greater the flow and the closer to the optimal value of efficiency under the same guide vane opening; By guide vane opening (S1 step output data) and water head Together determine: The greater, the greater the guide vane flow area, the greater the flow ; By water head And optimal water head Determination: when , the efficiency Reaches the maximum , the farther the deviation , the lower the efficiency; For real-time acquisition of unit speed data, reflecting the actual operation state of the unit, the higher the speed, the higher the frequency of torque work in unit time, and the torque determines the output power of the unit; The formula directly relates the output data of S1 step ( Indirectly determine ) and real-time acquisition of unit operation data ( , ), the calculated For single unit mechanical output characteristic data, it is the basis for subsequent cluster equivalent calculation.

[0070] (Three) cluster equivalent torque calculation (S3): from single machine torque to cluster equivalent aggregation Cluster equivalent torque calculation is the core step to realize "multi-machine equivalent to single machine", which is divided into dynamic weighted aggregation (S31) and cluster access simulation (S32). Through dynamic weight coefficient, each single machine data is associated to generate equivalent parameters meeting the overall characteristics of the cluster.

[0071] 1. Dynamic weighted aggregation: torque summation based on dynamic weight Formula: ; Symbol definition: Cluster equivalent torque, unit: N·m, The total number of units in the cluster, determined according to the actual simulation scene, such as Three Gorges Power Station cluster Take 32 units, Gezhouba Power Station cluster Take 21 units, the input scene configuration data, The first The dynamic weighting coefficient of the unit is dimensionless and ranges from 0 to 1. Calculated by step S4, For the first The output torque of the unit, in N·m, is calculated by step S2.

[0072] Data association: The cluster size data is the basis for determining the summation range, and is input based on the actual simulated hydropower unit cluster configuration; The dynamic weighting coefficient (output data of step S4) reflects the contribution of a single unit to the total torque of the cluster. Larger capacity units with faster dynamic responses have higher weighting coefficients. The larger; The single-machine torque data output in step S2 is the basis for aggregate calculation; The formula is passed Will The weighted summation of the individual torque data of each unit generates the equivalent torque of the entire cluster. This preserves the differences in dynamic characteristics among the units (through...) (This reflects the fact that) the computational workload of multiple units is simplified to that of a single equivalent unit, thereby reducing computational complexity.

[0073] 2. Cluster Access Simulation: Calculation of Equivalent Generator Speed formula: ; Symbol definition: This represents the equivalent generator speed change rate, in rad / s². The equivalent inertial time constant, in seconds, is derived from the inertial time constants of each individual machine. and The weighted calculation yields the following formula: , For the first The inertial time constant of the generator set, retrieved from the generator set parameter database, is typically 2~8s. The cluster equivalent moment, in N·m, is calculated by step S31. The equivalent damping coefficient, in N·m·s / rad, is derived from the damping coefficients of each individual unit. and The weighted calculation yields the following formula: , For the first The damping coefficient of the generator unit, retrieved from the unit parameter database, is typically between 0.5 and 2.0. This is the equivalent generator speed, in rad / s. The initial value is taken as the rated speed of the unit. rad / s, followed by Dynamic changes.

[0074] Data association: , For equivalent parameters, derived from the inherent parameters of a single machine ( , (Retrieved from the database) and dynamic weighting coefficients (The output data from step S4) is obtained through weighted calculation, directly reflecting the inertial and damping characteristics of the cluster. The larger the cluster size, the slower the speed change. The larger the cluster size, the faster the cluster rotation speed decays after being disturbed; The cluster equivalent torque data output in the dynamic weighted aggregation step is the "power source" that drives the change in the equivalent generator speed. The initial value is the rated speed (fixed input data), which is used as the equivalent speed. The real-time value is then obtained by integrating the differential equation and used directly as the core parameter of the power system simulation environment to reflect the response status of the cluster to the grid frequency. This formula will use cluster equivalent moments Convert to equivalent speed This enables the conversion from "mechanical characteristics" to "electrical characteristics," allowing the equivalent unit to be directly connected to existing power system simulation tools without requiring modifications to the simulation tools.

[0075] (iv) Dynamic weight coefficient calculation (S4): Adaptive weight allocation Dynamic weighting coefficients This is crucial for ensuring the accuracy of cluster equivalence. It requires combining the unit capacity ratio (static factor) and real-time dynamic response (dynamic factor) in the calculation to ensure that the weights can adapt to changes in unit operating conditions.

[0076] formula: ; Symbol definition: For the first The dynamic weighting coefficient of the unit is dimensionless and ranges from 0 to 1. The initial weights are unitless and calculated based on the proportion of unit capacity. The formula is as follows: , For the first The rated capacity of each generating unit, in MW, is retrieved from the unit parameter database. For example, the single unit capacity of the Three Gorges Power Station is 700MW. The adjustment coefficient is dimensionless and determined based on the power grid operation scenario: steady-state scenario. Disturbance scenarios Fault scenarios , Dynamic response factor, unitless, value range 0~1, calculated based on real-time response characteristics of the unit, formula is Wherein is the normalized speed change rate, is the normalized guide vane opening change rate.

[0077] Data association: Static weight, determined by the rated capacity of the unit (fixed inherent parameter, obtained from the database) and the total rated capacity of the cluster (calculated), to ensure that the larger the capacity of the unit, the greater the basic contribution to the cluster, which is the “static reference” of the weight; Dynamic response factor, calculated by normalizing the real-time collected speed change rate and guide vane opening change rate (both are real-time collected data, sampling interval 0.01s): ( is the maximum speed change rate of the unit, obtained from the database), ( is the maximum guide vane opening change rate of the unit, obtained from the database), reflecting the response speed of the unit to the power grid disturbance, the faster the response of the unit , the greater the value; Adjustment coefficient, dynamically assigned according to the power grid operating scenario (judged by real-time monitoring of the fluctuation range: steady state, disturbance, fault), to determine the influence degree of the dynamic response factor on the weight - the more severe the scenario (such as fault scenario), , the greater the adjustment effect, ; The formula ensures the static rationality of the weight through , and realizes the dynamic self-adaptation of the weight through , so that can reflect the capacity difference of the unit and adapt to the real-time operating condition, providing accurate weight support for the cluster equivalent calculation of step S3.

[0078] Three, realization of “decoupling aggregation” architecture To avoid the numerical instability problem caused by the direct coupling of multiple units and the power grid, the invention designs a “decoupling aggregation” architecture, which divides the simulation period through a time sequence control algorithm: 1. Period division: divide the total simulation period into single-machine decoupling simulation periods (70%) and cluster aggregation cycle (30%) 2. Decoupling simulation: In Within the unit, each unit independently executes steps S1-S2 to calculate the guide vane opening. With output torque Data exchange between units is not performed to avoid coupling interference; 3. Data Upload and Aggregation: After completion, all unit data is uploaded to the cluster control module. Execute steps S3-S4 internally to generate equivalent parameters ( , Access to the power system simulation environment; 4. Repeat the above process to achieve continuous and stable cluster simulation.

Claims

1. A method for simulating a cluster of hydroelectric generating units, characterized by: The method comprises the following steps of: S1, each governor calculates the guide vane opening degree according to the grid frequency deviation through PID control logic; S2, based on the nonlinear hydraulic turbine model, the output torque of a single unit is calculated by combining the unit head, flow, efficiency and speed parameters; S3, the dynamic weight coefficient is determined according to the unit capacity ratio and real-time dynamic response factor, the equivalent torque of the cluster is obtained by weighted aggregation, and the equivalent torque is input into the equivalent generator model, and the equivalent speed is calculated by combining the rotor motion equation; S4, the single machine dynamic characteristic independent calculation and cluster equivalent aggregation are realized through the "decoupling aggregation" architecture, which reduces the complexity of power system analysis and calculation while ensuring the simulation accuracy.

2. The water turbine unit cluster simulation method according to claim 1, characterized in that: Guide vane opening calculation: each unit governor calculates guide vane opening according to power grid frequency deviation by PID control logic The calculation formula is: ； wherein , is a reference frequency, is the actual grid frequency, , , are the proportional, integral and derivative coefficients of the governor, respectively; Unit torque calculation: based on the nonlinear water turbine model, combined with the unit parameter calculation single unit output torque , the formula is: ； wherein is the density of water, is the acceleration of gravity, is the head of the first is the head of the first is the flow of the first is the flow of the first is the efficiency of the first is the efficiency of the first is the rotational speed of the first is the rotational speed of the first The cluster equivalent torque calculation: Dynamic weighted aggregation: according to the dynamic weight coefficient of the unit , the equivalent moment of the cluster is calculated by the formula , is the total number of units in the cluster;​ Cluster access simulation: equivalent moment Input equivalent generator model, calculate equivalent generator speed combining with rotor motion equation The rotor motion equation is is equivalent inertia time constant, is equivalent damping coefficient);​ Dynamic weight coefficient calculation: through formula calculating dynamic weight coefficient wherein is the initial weight based on the proportion of unit capacity, is the real-time dynamic response factor of the unit, is the adjustment coefficient; The method adopts the "decoupling aggregation" architecture, first calculates the dynamic characteristics of each heterogeneous unit through S1-S2, then realizes the cluster equivalent aggregation through S3-S4, reduces the complexity of power system analysis and calculation while ensuring the simulation accuracy, and is suitable for power grid stability research and frequency regulation scene with large-scale water turbine units participating.

3. The method of claim 2, wherein: In step S1, the governor calculates the guide vane opening degree, which needs to be preprocessed first. The specific process includes: A1. Frequency acquisition: real-time acquisition of actual power grid frequency through high-frequency sampling module , the sampling frequency is set to 500Hz, and the continuous frequency data sequence is obtained ; A2. Filter processing: the collected frequency data is filtered by using Kalman filter algorithm to eliminate high-frequency interference signals, and the state equation is set as , is the process noise, and the observation equation is set as , is the observation value, is the observation noise; A3. Bias calculation: the filtered actual grid frequency is compared with the reference frequency Substitute , to get accurate frequency deviation value, and then input the PID control logic to calculate the guide vane opening .

4. The method of claim 2, wherein: In step S2, the parameters of the nonlinear hydraulic turbine model are determined by a working condition adaptation algorithm. The specific steps include: B1. Working condition classification: according to the unit operating head range, the unit working condition is divided into low head working condition, medium head working condition and high head working condition; B2. Parameter retrieval: For different working conditions, the corresponding characteristic parameters are retrieved from the unit parameter database, wherein the flow coefficient is taken as , the efficiency coefficient is taken as , the flow coefficient is taken as , the efficiency coefficient is taken as , the flow coefficient is taken as , and the efficiency coefficient is taken as for low, medium and high water head conditions respectively. B3. Parameter calculation: by formula Computer group flow , Corresponding to the flow coefficient of the working condition, by formula Computer group efficiency , Corresponding to the efficiency coefficient of the working condition, Optimal water head; B4. Moment calculation: the calculated , and real-time collected , substitute into the formula , to get the output torque of a single unit.

5. The method of claim 2, wherein: Dynamic weight coefficient in dynamic weighted aggregation Initial weight of The calculation needs to be implemented through a capacity proportion analysis algorithm, and the specific steps include:​ C1. Capacity collection: Collect the rated capacity of each unit in the cluster , forming a capacity data set ; C2. Total capacity calculation: by formula Calculate total rated capacity of cluster ; C3. Initial weight determination: According to the proportion of single unit capacity to total capacity, the initial weight of each unit is calculated by the formula , which ensures that the initial weight can reflect the basic contribution of unit capacity to the cluster output. ​ 6. The method of claim 2, wherein: Equivalent parameters of equivalent generator model in cluster access simulation 、 The calculation needs to be implemented through a parameter aggregation algorithm, and the specific steps include: D1. Single machine parameter acquisition: collect the inertia time constant of each unit in the cluster , damping coefficient ; D2. Weighted Calculation: Combining dynamic weight coefficients Through formula Calculate the equivalent inertial time constant Through formula Calculate the equivalent damping coefficient ; D3. Parameter verification: the calculated , is substituted into the rotor motion equation to simulate the rotor speed response under small grid disturbance. If the deviation of the response curve from the average rotor speed response curve of the units in the cluster is less than 5%, the parameters are determined to be valid; if the deviation is greater than or equal to 5%, the weight coefficients are adjusted again and the D2D3 steps are repeated until the parameters meet the accuracy requirements.

7. The method of claim 2, wherein: Dynamic response factor in dynamic weight coefficient calculation The calculation of the dynamic response factor needs to be implemented through a response characteristic analysis algorithm, and the specific steps include: E1. Response data collection: Collect the speed change rate of each unit in real time , the guide vane opening change rate , the sampling interval is set to 0.01s; E2. Normalization: This is achieved through the formula... Normalize the rate of change of rotational speed. The maximum rate of change of unit speed is expressed by the formula. Normalize the rate of change of guide vane opening. This represents the rate of change of the maximum guide vane opening of the unit. E3. Factor calculation: by weighted formula Calculate dynamic response factor Wherein the weight coefficients 0.6, 0.4 are determined according to the influence degree of speed response and guide vane response on the dynamic characteristics of the unit respectively.

8. The method of claim 2, wherein: Adjustment coefficient in dynamic weight coefficient calculation The determination of the adjustment coefficient needs to be implemented through a scene adaptation algorithm, and the specific steps include: F1. Scene classification: according to the grid operation scene, the simulation scene is divided into steady state scene, disturbance scene and fault scene; F2. Coefficient Assignment: Set and adjust coefficients for different scenarios. In the steady-state scenario Reduce the weight adjustment range to ensure simulation stability; under perturbation scenarios Appropriately increase the sensitivity of weight adjustment; in fault scenarios To maximize the effect of weight adjustment in order to adapt to drastic dynamic changes; F3. Scene switching: Real-time monitoring of power grid frequency fluctuation during simulation, automatic update of adjustment coefficient when scene switching , to ensure that the dynamic weight coefficient can adapt to the current power grid operation scene.

9. The method of claim 2, wherein: The realization of the "decoupling aggregation" architecture needs to be completed through a time sequence control algorithm. The specific steps include: G1. Period division: the total simulation period is divided into single-machine decoupling simulation periods and cluster aggregation periods. , and wherein 70%, 30%.​​ G2. Decoupled simulation: in S1-S2 steps are executed independently by each unit to calculate the guide vane opening and output torque , and no data interaction is performed between units to avoid coupling interference; G3. Data upload: At the end of the flight of all the crew , data is uploaded to the cluster control module; G4. Cluster aggregation: in S3-S4, complete cluster equivalent calculation, generate equivalent parameters and access the power system simulation environment; G5. Loop execution: repeat G2G4 steps to realize multi-period decoupling aggregation simulation, ensure continuous and stable simulation process.

10. The method of claim 2, wherein: For large-scale water turbine unit clusters, the units need to be grouped first through a cluster grouping algorithm, including: H1. Characteristic index extraction: Extract the key characteristic indexes of each unit, including the governor type, rated water head , rated capacity ; H2. Index normalization: by formula H2. Index normalization: by formula H2. Index normalization: by formula H3. Similarity calculation: the Euclidean distance formula is used to calculate the similarity of characteristics of any two units , 、 is the type identifier of the governor of the two units H4. Group determination: set a similarity threshold , the similarity of the units in the same group, and the equivalent parameters of each group are aggregated into the equivalent parameters of the entire cluster, reducing the computing pressure of the super large-scale cluster.

11. The method of claim 1, wherein: During the simulation process, the simulation accuracy needs to be verified in real time through an accuracy verification algorithm. The specific steps include: I1. Actual data collection: Real-time collection of actual total power of all units in the cluster through sensors , average speed of rotation ; I2. Simulation data acquisition: reading the equivalent power output from the cluster equivalent model output from the simulation system , equivalent rotational speed ; I3. Error calculation: by formula Calculate power error, by formula Calculate speed error; I4. Precision judgment: if and , it is determined that the simulation precision meets the requirements; if or , the dynamic weight coefficient is adjusted in step S4 , and steps S3-S4 are re-executed until the error meets the precision requirement.

Images